Новини

GitHub Copilot Code Review Expands with Agent Skills and MCP Integration

· GitHub Blog

GitHub has officially launched agent skills and Model Context Protocol (MCP) server support for GitHub Copilot code review, making these capabilities available to all Copilot Pro, Pro+, Business, and Enterprise subscribers. This represents a significant advancement in AI-powered code review technology, allowing development teams to integrate their existing tools, standards, and external data sources directly into the review process. The agent skills feature enables teams to customize Copilot's code review analysis by incorporating organization-specific coding standards and internal tools. Developers can create SKILL.md files within their repository's .github/skills directory to provide context and instructions tailored to their specific codebase or organizational requirements. This customization ensures that AI-generated feedback aligns with established team practices and coding conventions. MCP server integration represents another major enhancement, connecting Copilot code review to third-party platforms that teams already utilize, including issue tracking systems, documentation repositories, and service catalogs. This integration operates with read-only permissions to maintain security while pulling relevant context into the review process. Teams that previously configured MCP servers for Copilot's cloud agent will find their existing setups automatically apply to code review functionality. The general availability release includes improved attribution features, clearly indicating when review comments originate from agent skills or MCP context. This transparency helps developers understand the source of AI-generated feedback and track the effectiveness of their custom configurations. GitHub and Playwright MCP servers are enabled by default, providing immediate value for most development workflows. For organizations already using these features during the public preview phase, no configuration changes are required as existing setups continue functioning seamlessly. New adopters can configure MCP servers through repository settings and store authentication credentials securely using GitHub's secrets management system. This development positions GitHub Copilot as a more comprehensive code review solution, moving beyond generic AI assistance to provide contextually aware feedback that incorporates team-specific knowledge and external data sources. The integration of organizational tools and standards directly into the AI review process could significantly improve code quality and reduce the time developers spend on manual review tasks, while ensuring consistency with established practices across development teams.

GitHub has announced the general availability of agent skills and Model Context Protocol (MCP) server support for GitHub Copilot code review, expanding access to all Copilot Pro, Pro+, Business, and Enterprise users. This milestone follows a successful public preview period and represents a significant evolution in AI-powered code review capabilities.

The agent skills functionality allows development teams to extend Copilot's code review analysis with organization-specific context and coding standards. Teams can implement this by creating SKILL.md files within skill-specific subdirectories under .github/skills in their repositories. These files contain instructions and context that guide Copilot's review process, ensuring AI-generated feedback aligns with established team practices and organizational requirements.

MCP server integration provides another layer of customization by connecting Copilot code review to external platforms and tools that teams already use in their development workflows. This includes issue tracking systems, documentation platforms, and service catalogs. The integration maintains security through read-only access permissions, allowing Copilot to pull relevant context without modifying external systems.

A key advantage for existing users is seamless transition - teams that configured MCP servers during the public preview phase require no additional setup as their configurations automatically apply to the generally available version. Additionally, any MCP configurations previously established for Copilot's cloud agent functionality will work immediately with code review features.

The release includes enhanced attribution capabilities that clearly identify when review comments originate from agent skills or MCP context. This transparency helps developers understand the source of AI-generated feedback and evaluate the effectiveness of their custom configurations. GitHub and Playwright MCP servers are activated by default, providing immediate functionality for most development environments.

For new adopters, the setup process involves configuring MCP servers through repository settings under the Copilot section, with authentication tokens stored securely using GitHub's secrets and variables management system. The platform provides example MCP configurations to help teams get started quickly.

This advancement positions GitHub Copilot as a more sophisticated code review solution that goes beyond generic AI assistance. By incorporating team-specific knowledge, external data sources, and organizational standards directly into the review process, the platform can provide more relevant and actionable feedback. This contextual awareness could significantly improve code quality while reducing manual review overhead for development teams.

The integration represents a broader trend in AI development tools toward customization and context-awareness, allowing organizations to leverage AI capabilities while maintaining their specific workflows and standards. As development teams increasingly rely on AI assistance, the ability to customize these tools with organizational knowledge becomes crucial for maintaining code quality and consistency.

Посилання в статті